CH_TODO list:
==========


sample:

__enum (state, 2, (
  idle,
  startup,
  execute,
  shutdown
));

__struct (pt_cart_t,(
    (ch_bit8) x,
    (ch_bit8) y
));

__struct (pt_polar_t,(
    (ch_bit8) r,
    (ch_bit8) theta
));

__union (point_t,(
  (pt_cart_t) c,
  (pt_polar_t) p
));

*** add support for any datatype for mem/rom template T

*** Understanding conditional assignment ***
ch_bit4 a;
if (x > 0) {
  a[0-2] = 0;
  a[1] = 1; // override a[1]
  if (y > 0) {
    a[0] = 1; // override a[0]
    a[3] = 1;
  }
} else {
  a = 0;
}
*********************************************

X*** automatic dead code elimination
X*** add detection of duplicate assignments inside control block and outside (very important for bugs resolution)
X*** Assignment rule: by constructor assignent is ignored during check


*** prevent memory allocation inside conditionals 
*** fix memport reassignment
*** assigning a vector fron memory is broken (and reading the vector inside the branch)
*** the branch.owner concept is broken (a nested assignment to global would not prevent outer assignment) - use bitmask intead to check overrides
*** subscripts use inside control statement still broken









*** investigate alternative proxy implementation based on bitmask for simplicity
*** find a way to fix ch_reg such that it is initialized to itself by default (next -> this)
*** should consider gettng rid of defer initialization once auto dead code elimination is supported (will remove all nasty hacks in code where dst == nullptr is a worry)
*** address the issue why friend member still don't get implicit template constructor (e.g. ch_min, ch_sel, ...)


*** handling floating-point delays: (y = a * b) => ch_decoupled<ch_float32> object exporting ready/valid signals to be connected
*** complete float implementation (compare operations) 
*** make ch_mem store the enclosed type like ch_mem<ch_bit32, 10> 

 
*** use callstack api to save source file/line for undefined variables. check out GDB api (https://github.com/brasko/gdbwire) and libcwd (http://libcwd.sourceforge.net/index.html)
*** move all the math routines into the bitvector class, then accelerate the bitvector class using intel SSE/AVX intrinsics
*** add support for ch_reg implace operators ++(), --(), >>=(), <<=(), &=(), |=(), ^=(), +=(), -=(), *=(), \=()
*** test ch_shuffle
*** printing struct, union, enum values using reflection 
*** allow ch_println() to use __if and __switch 
*** handshking valid/ready signals

- implement ch_call(function, args...) using variadic template to generate modules binding 
- implement bitbase.clone() to enable fork assignments
- prevent assignments to yourself (e.g. a = a, a[0] = a[1])
- implement ch_make_device() to use compiler time binding
- add C++ user defined literals for _b, _h, _d, _o convertion to bit<N> and bus<N>
- create ../configure --prefix / make / make install scripts
- implement cash namespaces ch::logic, ch::system, ch::stl
- to_verilog() renaming module arguments
- investigate using std::bind and variadic functions for device creation
- add mutex and fifo primitives for TML modelling
- move memory functions into classes to support onetime initialization: ch_mem<N, W, AsyncR> mem(..); x = mem.read(addr); mem.write(en, addr, val);
- support ch_device_interface<type> for user to derive and implement emit()
- enforce type safety in CASH for explicit size casting
- experiment if/else using lambdas ch_when(a > 1, []{...}).when(b > 1, []{...}).otherwise([]{...})
- revisit the next() operator concept to make reg implicit
- check out usefull AVX shuffle api to support in CASH
- CASH conditionals: ch_when(?, :=).when(?, :=).otherwise(:=)
- global Reset device->reset() => manually reset all nodes in device to zero.
- FSM designs integration
- should generate high-level verilog
- refactor ch_tristate class => movech_bit<1>tion
- implement unit test framework like in MyHDL/chisel
- fix hierarchy support
- implement netlist export/import
- implement binary export/import
- implement AAL toolchain
- add direct initialization buffer support for ch_mem and ch_mem
- extend operators argument to support different bit size
- refactor ch_llrom
- add file initialization support to ch_llmem
- implement __union and __struct as part of aggregate.h
- implement ch_int and ch_unit classes
- implement plugin interface for attaching other simulators
- implement default simulator loging
- add source location information to user objects for debugging
- implement module functions generation and reuse
- implement new analysis api
- implement new netlist export/import
- implement GCC "#pragma cash module" to convert native functions to CASH module
- redesign schematic export to use new clock design
- accelerate simulation using a virtual machine
- support for high-level custom compoment plugins (with query cycle and gate count)
- support for persistence

Completed:
==========

- refactor ch_delay class => move to function
- refactor ch_wire class => all objects operate as wires before first assignment
- move ch_mem into llmem.h
- create cash.h header file
- adding native support for integers operations
- add CASH_ to all header files #define declarations
- implement ch::internal to cleanup api namespace
- keep all implementation specific headers private
- append ch_ to CASH public types and functions
- runtime assertion support for testing
- explicit clock implementation
- explicit clock simulation
- redesign concatenation support to use the comma operation
- add special assign operator '<=' for non-blocking assignments
- implement signal class
- implement efficient proxy classes for concatenate and slice operators
- headers cleanup/refactoring
- implement true assignment operator
- generic clock signal generation
- register enable/reset control
- optimize global states management (static objects => scoped objects)
- fix signals update during simulation (should only optimize read-only literals)
- redesign optimizer + remove all redundant loops
- redesign optimizer + optimize dead_code_removal calls
- optimize all loop to use auto& construct
- use C++ 11 std containers emplace api to reduce object construction overhead
- automic device instantiation
- new vcd simulator class
- implement ch_sel
- implement ch_case
- update llrom
- update llram
- fix memory implementation
- replace abort() with runtime assertion macro
- implement ch_pushcd, ch_popcd api (check out the opengl matrix api)
- implement new verilog generator

notes:
======

// debugging environment variables
export CASH_DEBUG_NODE=32
export CASH_DEBUG_LEVEL=2
export CASH_DUMP_AST=1

// infering muliple cycles operations
y = ch_pipeline(x, 2) = ch_delay(ch_delay(x));
z = ch_pipeline(x * y, 2); => z = ch_delay(mul1(x,y,ch_delay(mul0(x,y)));

// AOCL profiling report
aocl report jpeg_decoder.aocx profile.mon

// sync submodules
git submodule update --init --recursive

// Detecting leaks using valgrind
$ valgrind --leak-check=yes <app> <args...> > valgrind.log 2>&1
$ valgrind --leak-check=full --show-leak-kinds=all --track-origins=yes <app> <args...> > valgrind.log 2>&1

// CPU Profiler
$ valgrind --tool=callgrind <app> <args...>
$ kcachegrind

// google perf tools

$ CPUPROFILE_FREQUENCY=100000 CPUPROFILE=profiler.log ./aes.out -e
$ google-pprof --callgrind ./aes.out profiler.log > profiler.callgrind
$ kcachegrind profiler.callgrin

// cmake build
cmake -DCMAKE_BUILD_TYPE=Release .. && make -j`nproc` all test
cmake -DCMAKE_BUILD_TYPE=Debug .. && make -j`nproc` all test
cmake -DCMAKE_BUILD_TYPE=Release -DJIT=LIBJIT .. && make -j`nproc` all test
cmake -DCMAKE_BUILD_TYPE=Debug -DJIT=LIBJIT .. && make -j`nproc` all test
cmake -DCMAKE_BUILD_TYPE=Debug -DCODECOV=ON .. && make -j`nproc` all test && ../codecov.sh .

// dump assembly
objdump -d a.out > a.asm

// debugging C macros
g++ -std=c++17 -O0 -g -Wall -Wextra -Werror -E -dD dogfood.cpp > dogfood.log
clang-format-5.0 dogfood.log -i

// Detecting Cycles in CASH?
// cycles only occurs when combinatorial nets eval() function is entered a second time without having exited.
bool nandimpl::eval(ch_tick t) {
  if (tick_ != t) {
    if (tick_ > t) {
        CASH_ABBORT("found cycle");
	}
    tick_ = t + 1;
    cval_ = !(this->src(0).eval(t) && this->src(1).eval(t));
    tick_  = t;
  }
  return cval_;
}

//
// alternative CASH programming model
// enable realtime debugging in behavioral mode
//
template <unsigned N>
void Counter(ch_bit<N>& out) {
  ontick([&]() {
    out = out + 0x1;
  });
}

// user-defined literals for binary numbers
constexpr unsigned long long Scale(unsigned long long x, const char* s) {
    return (!*s ? x : Scale(10*x + ((unsigned long long)*s)-((unsigned long long)'0'), s+1));
} 

constexpr unsigned long long ToScaledBinary(unsigned long long x, unsigned long long scale, const char* s) {
    return (!*s 
                ? x 
                : ( *s == 'e' || *s == 'E'  
                        ? (x << Scale(0,s+1)) // put the digits in the right position
                        : ToScaledBinary(x + x + (*s =='1'? 1 : 0), scale,  s+1)));
}
std::cout << std::hex << "0x" << 1111111111_b << std::endl;    
std::cout << std::hex << "0x" << 1111111111e16_b << std::endl;

//--
ch_device<Adder<2>> my_adder;
__tie(out, cout) = my_adder.bind(lhs, rhs, cin);

template <typename From, typename To>
class is_explicitly_convertible {
private:
  template <size_t N>
  struct dummy_t {};

  template <typename F, typename T>
  static std::true_type check(dummy_t<sizeof(static_cast<T>(std::declval<F>()))>*);

  template <typename F, typename T>
  static std::false_type check(...);

public:
  static constexpr bool value = decltype(check<From, To>(nullptr))::value;
};

#include <iostream>

template <typename Child>
struct Base
{
    void interface()
    {
        static_cast<Child*>(this)->implementation();
    }
};

struct Derived : Base<Derived>
{
    void implementation()
    {
        cerr << "Derived implementation\n";
    }
};

int main()
{
    Derived d;
    d.interface();  // Prints "Derived implementation"
}

http://stackoverflow.com/questions/25513380/c-style-callbacks-in-c11
http://en.cppreference.com/w/cpp/thread/packaged_task
http://stackoverflow.com/questions/38561242/struct-to-from-stdtuple-conversion

clang++-9 -std=c++17 -I../include -lcash -L$CASH_HOME/build/lib -Xclang -load -Xclang lib/libcashpp.so -Xclang -add-plugin -Xclang cash-pp -Xclang -plugin-arg-cash-pp -Xclang debug -Xclang -plugin-arg-cash-pp -Xclang dump-ast ../examples/adder.cpp > test.log 2>&1
clang++-9 -std=c++17 -I../include -Xclang -load -Xclang lib/libcashpp.so -Xclang -add-plugin -Xclang cash-pp -Xclang -plugin-arg-cash-pp -Xclang debug -Xclang -plugin-arg-cash-pp -Xclang dump-ast ../clang/test.cpp > test.log 2>&1

bundle exec jekyll serve --trace